Results From CW Stabilized Link . - Stanford University

Results from CW stabilized link timingdistribution at LCLSRussell WilcoxGang HuangLarry DoolittleJohn ByrdJohn Staples

Application and concepts Requirement: synchronize multiple lasers to a reference signal– Original spec was 100fs RMS maintained over 8 hoursEmphasis on reliability, as downtime is costlyConcept: synchronize lasers to RF signal transmitted on fiber– Well-established high harmonic laser locking technique We demonstrated 15fs synchronization of two lasers at 2.5GHz– RF-over-fiber is cable TV technology All fiber telecom parts for reliability and cost Standard telecom fiber– Uses LBL-developed low noise digital phase detector 0.01 degree phase sensitivity (10fs at 3GHz)Optical interferometer to senses fiber delay change– High temporal resolution– Low noise heterodyne interferometerInterferometer reports to digital phase detector, which then appliescorrection to received RF– No mechanical time delay adjusters

Environmental perturbations offiber, cable, laser MaterialCoeff. of delay per deg C delay for 1m, 1deg.CSteel15 x 10 -650fsAluminum22 x 10 -672fsFiber8 x 10 -640fsCoax, teflon-85 x 10 -6-425fsCoax, air heliax-10 x 10 -6-50fsAir (thermal)-3 x 10 -6-10fsAir (pressure)2 x 10 -6 / 10 millibars7fs / 10mbarAir (humidity)4 x 10 -6 / 10%RH13fs / 10%RHthermal coefficient of index is the main driver for fiber

Schematic of one linkreceivertransmitterCWlaserAMd2FRMRblock ωRFfiber 10.01Cfiber 20.01CFSωFSFRMd1RF phasedetect,correctopticaldelaysensingFRM is Faraday rotator mirror (ends of the Michelson interferometer)FS is optical frequency shifterCW laser is absolutely stabilizedTransmitted RF frequency is 2856 MHzDetection of fringes is at receiverSignal paths not actively stabilized are temperature controlled

Schematic of one linkreceivertransmitterCWlaserAMd2FRMRblock ωRFfiber 10.01Cfiber 20.01CFSωFSFRMd1RF phasedetect,correctopticaldelaysensingFRM is Faraday rotator mirror (ends of the Michelson interferometer)FS is optical frequency shifterCW laser is absolutely stabilizedTransmitted RF frequency is 2856 MHzDetection of fringes is at receiverSignal paths not actively stabilized are temperature controlled

Schematic of one linkreceivertransmitterCWlaserAMd2FRMRblock ωRFfiber 10.01Cfiber 20.01CFSωFSFRMd1RF phasedetect,correctopticaldelaysensingFRM is Faraday rotator mirror (ends of the Michelson interferometer)FS is optical frequency shifterCW laser is absolutely stabilizedTransmitted RF frequency is 2856 MHzDetection of fringes is at receiverSignal paths not actively stabilized are temperature controlled

Expansion to multiple channelsref.armfreq.shifterLLRF controller100MHz beatRF inCWfiberlaserAMref.armfreq.shifterLLRF controllerfreq.shifterLLRF controllerRbfreq.lockerref.arm Since all processing is at reciever, a multi-channel transmitter is notcomplex– 32-channel amp/splitter/ref arm fits in 8U (14”) high rack chassis

Two-channel test interferometer2kmfreq.shifter 50MHz2kmCWfiberlaser100MHzbeat100MHz beatRF controller2mphase data2m freq.shifter-50MHzThis is an out-of-loop test to see if the interferometers are workingAlso, it’s a measurement of the actual drift and noiseWe installed this in LCLS, and measured tunnel and gallery 2km fibers

Test interferometer resultsunlockeddelay variationlockeddifferential errorfrequency, Hz time, hoursTranslating phase error on the 100MHz beat note to 200THz optical, theintegrated jitter is 0.25fs (assuming perfect wavelength stability)With 2ns total correction, the average drift is 3fs (less than one wave) per day– We later found the monotonic drift was a computation artefact

Receiver functions are implementedby a digital phase detector 14 bit DACs125MHz sample rateIt controls both the interferometer and RFphase locked loops

Phase detector stability testsignalcalibration (common mode)2856MHzreference Blue area is temperature stabilizedSignal paths to digitizer are not delay stable– We are measuring the phase difference between signal and reference– The calibration signal presents a common mode signal to both paths,so that differential delay changes can be subtracted out

Phase stability test results 24 hours, 125kHz bandwidth, 2856MHz inputUncorrected differential temporal error, 140fs RMSCorrected differential temporal error, 15fs RMSWe are close to the theoretical limit, given the the noise figure of thecomponents

We tested a dual channel system2km0.01C0.01C1560nmCWfiberlaserRbfreq.locker RF inAMRF inref.arm 50MHzref.arm 50MHzRF inRF aldelaysensingOpt. Lett. 34, 3050 (2009)Measurement of the differential phase variation between two stabilized links

2.2km 200mAllan deviationdelay error, femtosecondsDual-channel resultstime, hours1kHz bandwidthFor 2.2km, 19fs RMS over 60 hoursFor 200m, 8.4fs RMS over 20 hours2-hour variation is room temperature2km datatime, seconds

We correct for group versus phase delay Group delay is not equal to phase delay, due to dispersionng n ϖ dndϖand alsodngdT dndTA temperature dependent Sellmeier equation was fitted to previous data byGhosh et al (IEEE JLT 12, 1338)B (T )D (T ) ng (T ) n(T ) 1%22(1()/)(1()/)λλ CT ETWe correct for group delay by adding RF delay proportional to the opticaln 2 A(T ) correctionearly testwith and withoutfeedforwardcorrection:group error asfiber heats1.6% correction added36fsRMS

We determine the additional correctionby adjusting to minimize error in situRF infiberundertesttuneTXshortfiber RX1RX2Adjust feedforward correction until error is minimizedWe don’t find significant changes to this factor– Tested mainly on multi-fiber SMF cableVCO

Power-to-phase conversion inthe photodiode is not a 2GHznetworkanalyzerpowermeter-0.4data and fit fortwo diodes-0.6-0.8-10246810optical power, mW Near saturation, high density of photocarriers screens applied field– Carriers are not swept out, response is slowed– Why it’s not monotonic is unclear, but it’s useful /- 10% power variation around zero slope point causes 10fs time shiftIn practice, power is stable to this degree and we don’t have to regulate– This is an option12

Manufacturer knows about this,is improving diodesJoshi and Datta, IEEE Phot. Tech. Lett. 21, 1360 (2009) Their results were for pulses, 1GHz harmonicWe need to test these with modulated CW, which has much smaller effectAt least we don’t have to worry the effect will get worse– New zero slope point is OK, reduces power requirements

LCLS timing scheme 150mphasecavityundulatornear-end halle-experimentarrivaltimemonitor476laser2856 and 68receiverX6φ2856φreceivertiming information4762856X64.25 We sync to bunch arrival time monitorThe laser is treated as a VCOdivider120HztriggersendermodulatorCW laser

2-channel, out-of-loop, in situ testRF in300mTX2856MHz 27fs RMS in 125kHz BW16fs RMS in 1kHz BWDrift is due to short cablebetween receivers, roomtemperaturefiber5mRX1RX2tuneVCO

Laser locking configurationfrep out (68)RF inRF phasedetect RF out(cal.)andcorrectRF inreference 2856MHzand 4.25MHz X6harmonicout (476)sync headlaserpiezo68 MHzpulse trainPhase compare at 2856MHzSync first to 68MHz to remove “bucket ambiguity”Works better than the commercial lockboxNew arrangement uses faster diode, eliminates X6 multiplier

In-loop resultsRF control error signal125kHz BW (gray): 31fs RMS1kHz BW (black): 8fs RMS Laser control error signal125kHz BW (gray): 120fs RMS1kHz BW (black): 25fs RMSImprovements to the laser should decrease high frequency noise– Acoustic and vibration isolation– Lower noise pump

The transmitter fits in a standard rack VCO Receiver(for laser)Splitter Diagnostic Amplifier Modulator Wavelengthlocker CW laser

Conclusions, future work We have demonstrated a laser-to-RF sync system in an FEL– 16fs between two RF channels, 25fs laser loop error (1kHz)– Used reliably for experiments (as reported earlier)Easily manufacturable, expandable– First commercially produced subsystems being tested– LCLS is engineering next version, will be making 8 channels soon,upgrading transmitter to 16 channel capabilityFuture work– Improve laser control– Better synchronization measurements– Try higher frequencies

FRM is Faraday rotator mirror (ends of the Michelson interferometer) FS is optical frequency shifter CW laser is absolutely stabilized Transmitted RF frequency is 2856 MHz Detection of fringes is at receiver Signal paths not actively stabilized are temperature controlled Rb lock 0.01C AM CW laser 0.01C FS RF phase detect .